Mechanisms of resistance associated with the inhibition of the dehydration step of FAS-II in Mycobacterium tuberculosis

Abstract

Isoxyl (ISO) and thiacetazone (TAC) are two antitubercular prodrugs that abolish mycolic acid biosynthesis and kill Mycobacterium tuberculosis (Mtb) through the inhibition of the essential FAS-II dehydratase, HadAB. While mutations preventing ISO and TAC from either being converted to their active form or from covalently modifying their target are the most frequent spontaneous mutations associated with high-level resistance to both drugs, the molecular mechanisms underlying the high-level ISO and TAC resistance of Mtb strains harboring missense mutations in the second, non-essential, FAS-II dehydratase HadBC remained unexplained. Using a combination of genetic, biochemical and biophysical approaches and molecular dynamics simulation, we here show that all four reported resistance mutations in the HadC subunit of HadBC alter the stability and/or specific activity of the enzyme, allowing it in two cases (HadBC^V85I and HadBC^K157R) to compensate for a deficiency in HadAB in whole Mtb bacilli. Analysis of the mycolic acid profiles of Mtb strains expressing the mutated forms of HadC further points to alterations in the activity of the mycolic acid biosynthetic complex and suggests an additional contributing resistance mechanism whereby HadC mutations may reduce the accessibility of HadAB to ISO and TAC. Collectively, our results highlight the importance of developing optimized inhibitors of the dehydration step of FAS-II capable of inhibiting both dehydratases simultaneously, a goal that may be achievable given the structural resemblance of the two enzymes and their reliance on the same catalytic subunit, HadB.

Graphical Abstract

Mycolic acids are essential constituents of the outer membrane of all mycobacteria^1–2. The biosynthesis of these long-chain (C₆₀-C₉₀) α-branched, β-hydroxylated, fatty acids is a validated target for drug discovery as illustrated by the successful use of a number of past and present antitubercular agents such as isoniazid (INH), ethionamide (ETH), thiacetazone (TAC), and isoxyl (ISO)³. While the biosynthesis and export of mycolic acids involve the concerted action of over 30 proteins^4–5, it is noteworthy that inhibitors of this pathway that have made it to the clinic thus far only target two catalytic steps, both of them located within the type II Fatty Acid Synthase (FAS-II) required to iteratively elongate the meromycolate chain of mycolic acids. FAS-II is composed of four dissociable sets of enzymes: The reductase MabA, the dehydratases HadAB and HadBC, the enoyl-ACP reductase InhA, and the condensases KasA and KasB. INH and ETH inhibit the enoyl-ACP reductase InhA, whereas ISO and TAC target the essential dehydratase HadAB^4,6. This observation together with the recent report that two inhibitors targeting different steps of FAS-II used in combination not only display synergistic lethality in vitro and in vivo but also have the ability to kill persistent Mycobacterium tuberculosis (Mtb) bacilli⁷ provides a new incentive to develop additional inhibitors of the FAS-II elongation system. In this context, work in our laboratory has focused on elucidating the molecular mechanisms of susceptibility and resistance of Mtb to ISO and TAC with the goal of developing novel, more potent, inhibitors of the same catalytic step of the FAS-II cycle displaying reduced resistance mutation frequencies and more desirable pharmacological features.

ISO and TAC are thiourea prodrugs that require metabolic oxidation of their thiocarbonyl moiety by the flavin-dependent monooxygenase EthA to become bactericidal. The activated forms of the drugs inactivate the HadAB dehydratase by forming a covalent adduct with the Cys61 residue of the HadA subunit of the enzyme⁶. Consistent with this mode of action, C61G and C61S mutations in HadA and loss-of-function mutations in EthA are the most frequent spontaneous mutations associated with high-level resistance to ISO and TAC^8–13. Other mutations reported in ISO and TAC spontaneous resistant mutants of Mtb and Mycobacterium kansasii are less well characterized and not as readily explained^10,12–14. Missense and frameshift mutations in the methyltransferases MmaA2 and MmaA4 which are responsible for the introduction of functional groups in the meromycolate chain of mycolic acids confer resistance to ISO and TAC but the available evidence suggests these proteins are neither direct targets nor activators of the prodrugs^10–11. In light of the fact that both MmaA2 and MmaA4 physically interact with HadAB via the HadA subunit, we thus proposed earlier that resistance associated with mutations in these enzymes may result from conformational changes in the FAS-II multi-protein complex^15–18 limiting the drugs’ access to HadAB¹¹.

Similarly, the molecular mechanisms underlying the high-level ISO and TAC resistance associated with missense mutations in the HadC subunit of the second dehydratase of FAS-II, HadBC, are not directly intuitive^10–13. HadAB and HadBC are two independent heterodimeric enzymes sharing the same catalytic subunit, HadB. While HadA and HadC do not carry any catalytic residue, they are believed to play analogous functions in the stabilization of long fatty acyl substrates¹⁹. Evidence based on enzyme assays, physical interactions between the dehydratases and other FAS-II enzymes, and the phenotypic analysis of an Mtb knock-out mutant deficient in HadC points to HadAB dehydrating short meromycolyl precursors during the early stages of their elongation by FAS-II and HadBC acting at later stages, on longer fatty acyl substrates^17,19–20. In line with the late involvement of HadBC in the pathway, HadBC is a dispensable enzyme in vitro although its presence is required for normal mycolic acid production and full virulence in a murine model of tuberculosis infection^19–20. HadAB, in contrast, is essential for growth^19,21–23.

Because of the importance of understanding the mechanisms of resistance of Mtb to HadAB inhibition as we develop optimized inhibitors of this critical FAS-II enzyme, the present study was undertaken with the goal of elucidating the molecular mechanisms underlying the high-level resistance of HadBC mutants to ISO and TAC.

RESULTS

Generation of isogenic Mtb strains with resistance mutations in hadA and hadC –

Prior to gaining insight into the mechanisms underlying the resistance of hadC mutants to ISO and TAC, we first verified that the four hadC mutations identified in spontaneous-resistant mutants, HadC^V85I, HadC^T123A, HadC^A151V and HadC^K157R actually increased the resistance of Mtb to both drugs when introduced by recombineering in a susceptible Mtb H37Rv strain. As a control, a HadA^C61S mutant was also generated using the same approach. MIC determinations indicated a greater than 16-fold increase in the MICs of ISO and TAC against all mutants relative to the wild-type (WT) parent strain [Table 1].

Table 1: MIC of ISO and TAC against Mtb strains expressing the different hadA and hadC resistance mutations.

MIC values are in μg/mL. MIC determinations were performed on three independent culture batches.

Strains	ISO	TAC	INH
Mtb H37Rv mc26206 (WT)	1.25	0.62	0.05–0.1
Mtb HadA C61S	> 20	> 20	0.05
Mtb HadC V85I	> 20	> 20	0.05
Mtb HadC A151V	> 20	> 20	0.05
Mtb HadC T123A	> 20	> 20	0.05
Mtb HadC K157R	> 20	> 20	0.05

The protective effect of the missense HadC mutations further reflected in the growth curves of the Mtb mutants exposed to 8- (ISO) to 16-X (TAC) the MIC concentrations of the drugs although, using this method, it became apparent that the level of protection conferred by the mutations in HadC was less than that conferred by the C61S mutation in HadA [Fig. 1]. While WT Mtb failed to grow in the presence of the drugs, the HadA^C61S mutant was unaffected, and the four HadC mutants displayed an intermediate growth phenotype characterized by a ~7 to 24% decrease in replication rate relative to the DMSO-treated controls. The HadA and HadC mutants otherwise replicated at a similar rate as the WT parent strain in the absence of drug in the medium.

Figure 1: Growth rates of wild-type Mtb and its isogenic hadA and hadC resistant mutants in the presence and absence of ISO and TAC.

Growth curves were carried out at 37°C in 7H9-OADC-Tyloxapol medium containing 10 μg/mL ISO (orange line), 10 μg/mL TAC (green line) or no drug (2% DMSO control; black line).

The mutations in HadC partially protected Mtb against the effect of ISO on mycolic acid biosynthesis as evidenced by the metabolic labeling of WT and mutant bacilli with [¹⁴C]-acetate [Fig. 2A]. Consistent with the growth curves presented in Fig. 1, the HadA^C61S mutant was the one whose de novo mycolic acid synthesis was the least affected by the drug.

Figure 2: Effect of ISO treatment on the mycolic acid composition of Mtb expressing different mutated variants of hadA and hadC.

(A) Effect of ISO on mycolic acid biosynthesis. WT Mtb and isogenic mutants expressing the C61S variant of HadA or the V85I, T123A, A151V and K157R variants of HadC grown in 7H9-OADC-tyloxapol at 37°C were treated with either no drug, 7 μg/mL ISO or 14 μg/mL ISO and metabolically labeled with [1,2-¹⁴C]acetic acid for 13 h. The same total cpm of [¹⁴C]-acetate-labeled fatty acid and mycolic acid methyl esters (FAMEs and MAMEs) from treated and untreated cells were analyzed by TLC in the solvent system [n-hexanes:ethyl acetate 95:5; by vol.; three developments] and revealed by PhosphorImaging. The amount of radioactivity incorporated in mycolic acids (including alpha-, methoxy- and keto-mycolates) in the untreated cells and the cells treated with 7 μg/mL ISO was semi-quantified using a PhosphorImager, and the results (expressed as a percentage decrease relative to the value measured in the untreated control) are presented as a histogram.

(B) GC/MS analysis of the trimethysilylated fatty acids prepared from the same untreated and ISO-treated (10 μg/mL for 24 h) WT and mutant Mtb strains as above. 16-hydroxy palmitic acid was used as an internal standard for the quantification of 3-hydroxy fatty acids in the samples. Results are expressed as fold-increases in drug-treated cells relative to the untreated control and are representative of two independent experiments (biological duplicates). The dashed line denotes a ratio of 1.

The partial level of protection afforded by the missense mutations in HadC also reflected in the impact of ISO treatment on the build-up of 3-hydroxy-meromycolate precursors. Consistent with earlier findings¹¹, exposure of WT Mtb to the drug led to a significant (2- to more than 20-fold) increase in the amount of 3-hydroxy-C₂₀ to C₂₈ fatty acids in the cells [Fig. 2B]. No such dramatic increases were detected in the HadA^C61S mutant whose individual 3-hydroxy-fatty acid contents remained within 0.5 to 1.1-fold of the levels detected in the untreated cells. In comparison, the HadC mutants all displayed intermediate phenotypes with 3- to up to 10-fold increases in 3-hydroxy fatty acids 20 to 32 carbons in length, whose nature depended on the specific mutant. Of the HadC mutants, HadC^K157R was the one for which the build-up of 3-hydroxy-meromycolate precursors was overall the least marked.

Effect of missense resistance mutations in HadC on mycolic acid biosynthesis in Mtb –

To determine whether the resistance mutations found in HadC had any impact on the biosynthesis of mycolic acids, we first labeled untreated Mtb WT and HadC mutants with [¹⁴C]-acetate and compared their mycolic acid methyl ester (MAME) profiles by TLC. The only distinctive MAME pattern noted by this method was that of the HadC^T123A mutant [Fig. 2A]. Compared to other strains, this mutant presented significant and reproducible ~ 33% and 66% decreases in methoxy- and keto-forms of mycolates, respectively, compensated by an increase in alpha-mycolates [Fig. 3A].

Figure 3: Mycolic acid composition of Mtb expressing different mutated variants of hadC.

(A) WT Mtb and isogenic mutants expressing the V85I, T123A, A151V and K157R variants of HadC were grown in 7H9-OADC-tyloxapol at 37°C and metabolically labeled with [1,2-¹⁴C]acetic acid. Upon preparation of their MAMEs, samples were run on TLC as in Fig. 2A and the relative amounts of radiolabel contained in the alpha-, methoxy- and keto-mycolates of each strain were quantified by PhosphorImaging. The results shown are averages ± standard deviations of two independent experiments (biological replicates). Asterisks denote statistically significant differences between WT and mutated HadBC variant for each mycolic acid species pursuant to the Student’s t-test (P < 0.05).

(B) Mycolic acids extracted from the cells of WT Mtb and isogenic mutants expressing the V85I, T123A, A151V and K157R variants of HadC were analyzed by LC/MS for the presence of an additional unsaturation, a modification specifically linked to hadC loss-of-function mutations²⁰. The relative amounts of alpha- (upper panel) and keto-mycolic acids (lower panel) with an additional unsaturation are shown. The labels above the columns stand for the number of carbon atoms in the mycolic acids.

LC/MS analyses further revealed a 1.2 to 9.5-fold increase in alpha- and keto mycolates harboring an additional unsaturation in the HadC^T123A mutant relative to the WT parent strain [Fig. 3B]. Although not as marked as in the HadC^T123A strain, increases in the same forms of alpha and keto mycolates were also noted in the other three HadC mutants [Fig. 3B]. Interestingly, these changes in degree of unsaturation did not impact the methoxy mycolates of any of the mutants (data not shown). The length of the hydrocarbon chain of the mycolates prepared from the four mutants otherwise did not differ from that of the WT strain (data not shown). While an increase in the degree of unsaturation of oxygenated mycolates and, to a lesser extent, alpha-mycolates was previously reported in Mtb strains completely deficient in HadBC activity, these changes impacted both the keto and methoxy forms of mycolates and were accompanied by a total loss of trans-cyclopropanated rings²⁰. The mycolic acid composition of the four HadC mutants thus only partially phenocopied that of HadC-deficient strains and points to the activity of some of the enzymes functionalizing the meromycolate chain of mycolic acids with double bonds and oxygenated functions being altered in these resistant strains.

Resistance mutations in HadC have no impact on the expression level of the hadABC operon –

Because the overexpression of the hadABC operon was shown to confer high-level resistance to ISO and TAC in Mtb, we next compared by qRT-PCR the expression levels of hadAB and hadBC in the four HadC resistant mutants to those measured in the WT parent Mtb strain. Consistent with earlier results¹³, the HadA^C61S and HadC^V85I resistant mutants expressed the hadA and hadC genes at similar levels as WT Mtb [Fig. 4]. Less than 2-fold increases in the levels of hadA and hadC transcripts were also observed in the HadC^T123A, HadC^A151V and HadC^K157R mutants [Fig. 4] thereby excluding an increase in the level of expression of the hadABC operon in the mutants as the cause of their high-level resistance to the pro-drugs.

Figure 4: Expression levels of the hadABC operon in the drug-resistant HadC mutants.

WT Mtb, isogenic mutants expressing the C61S variant of HadA or the V85I, T123A, A151V and K157R variants of HadC, and the recombinant Mtb strains overexpressing hadABC under control of the hsp60 promoter from the replicative plasmid pVV16¹¹ were grown in 7H9-OADC-tyloxapol at 37°C to mid-log phase. Total RNA was isolated from three independent cultures and the levels of hadA and hadC transcripts relative to those of the housekeeping sigA gene were determined by quantitative reverse transcription-PCR. Ratios of hadA/sigA or hadC/sigA mRNA are means ± standard deviations (n = 3 RNA extractions and qRT-PCR reactions). Asterisks denote statistically significant differences between WT and recombinant strains pursuant to the Student’s t-test (P < 0.05).

Effect of resistance mutations on the dehydratase activity of HadBC –

The observation by Sacco et al.¹⁹ that purified HadBC may process some of the same short (C₁₂-C₂₀) substrates as HadAB in vitro (albeit ~ 100-fold less efficiently than the latter enzyme) suggests that some level of redundancy between the two enzymes may exist in whole cells, which the presence of specific mutations in HadC may exacerbate. Preliminary evidence indicating that the substrate specificity of at least two out the four HadC variants may be altered was provided by structural modeling and molecular dynamics simulations. The location of the mutations on a homology model of HadBC is shown on Fig. 5A. Molecular dynamics simulation (one microsecond production run) on the WT enzyme and the K157R mutant [Fig. S1] indicated different modes of interaction for the K157 and R157 residues. While the arginine interacts with the flexible loop between αHD and β2²⁴ via M79 and V77, the lysine residue of the WT enzyme which extends into the solvent does not interact with the loop [Fig. 5B]. This difference in interaction is thought to be caused by the different length of the side chains of arginine and lysine. T81, located on the same loop and situated at the entrance to the channel, is displaced in the mutant enzyme as a result, and the size and shape of the channel is altered. The calculated volume of the channel at 1 μs in the WT is 839 ų while that of the K157R mutant is 1005 ų. The change in shape of the entrance to the channel is shown in [Fig. 5D] and [Fig. 5E] and is more rounded in the case of the K157R mutant. The increased volume and altered shape do not directly explain the altered activity of the mutant enzyme, but they do indicate the presence of structural changes in the mutant enzyme that could affect substrate recognition.

Figure 5: Structure of the HadBC enzyme showing the position of the resistance mutations and effect of mutations on shape of channel entrance.

(A) Homology model of HadBC generated from the HadAB structure²⁴ (PDB:4RLT), showing the positions of the HadC mutations (V85, A151, K157, T123) and the HadB catalytic residues (H41 and D36). The flexible loop is located between the αHD alpha helix and the β2 beta sheet. Subunits B and C of the enzyme are marked. (B) Overlaid conformations of the WT and the K157R mutant, showing a difference in the position of the T81 residue at the channel entrance, stemming from differences in interaction between the flexible loop and the residue at position 157. In the mutant, R157 interacts with the loop whereas in the WT HadC subunit, K157 does not. (C) Posterior view of the complex, showing the position of the V85 residue and the change in shape of the entrance (in blue) upon mutation to I85. (D) and (E) Change in shape of the channel entrance at 1μs of the production MD simulation. (D) Shape of the channel entrance of the WT enzyme. (E) Shape of the channel entrance of the K157R mutant.

Unlike A151 and K157, V85 is conserved in HadA and HadC [Fig. S2]. This residue is positioned at the posterior end of the channel. The difference in amino acid side chain alters the shape of the channel opening [Fig. 5C], possibly facilitating the binding of smaller substrates. The two remaining mutations, T123A and A151V, have no readily predictable impact on the active site of HadBC. T123A, lines the cavity that is perpendicular to the plane of the fatty acid binding channel [Fig. S3]. The cavity has been reported to serve as the binding site of small molecule inhibitors in HadAB²⁴, including ISO and TAC⁶. This mutation alters the physicochemical environment of the pocket as a polar, hydrogen bond-donating, side chain is mutated to a neutral, more hydrophobic residue.

The hypothesis that the HadBC mutants may at least partially compensate for a deficiency in HadAB by processing shorter fatty acyl substrates than the WT HadBC enzyme was first tested in vitro. To this end, the HadBC WT and mutated variants were produced and purified from E. coli and tested for activity on trans-2-hexadecenoyl-CoA (trans-2-C_16:1-CoA) using a previously described spectrophotometric assay¹⁹. Enzymes of the same dehydratase family as HadAB and HadBC are indeed known to preferentially catalyze the hydration of enoyl derivatives rather than the reverse dehydration reaction when isolated from their enzymatic complex¹⁹. Kinetic experiments revealed a significant 1.8- to 3-fold increase in the specific activity of the HadC^T123A, HadC^K157R and HadC^V85I variants relative to WT HadBC toward trans-2-C_16:1-CoA [Fig. 6A]. The activities of these three variants, however, remained about 40 to 70-fold less than that measured for the HadAB enzyme under the conditions of our assay. Shorter trans-2-enoyl-CoA substrates (C₈-C₁₂) were also tested but the very low initial velocities measured with all enzyme variants for these substrates (~ 5 times lower than that measured with the C₁₆ substrate) rendered comparisons unreliable.

Figure 6: Comparative enzymatic activity and thermal stability of the wild-type and mutated HadBC proteins.

(A) Comparison of the specific activities of HadAB (13 nM) and the WT and mutated variants of HadBC (270 nM of each protein) using trans-2-C_16:1-CoA (25 μM) as the substrate. Data are means ± standard deviations of triplicate assays and are representative of three independent experiments using different batches of purified proteins. Asterisks denote statistically significant differences between WT and mutated HadBC variant pursuant to the Student’s t-test (P < 0.05).

(B) Comparison of thermal stability profiles for HadBC WT and mutant proteins. Average inflection temperatures ± standard errors were determined using NanoDSF from triplicate runs. They show the K157R and V85I mutants to be more stable than WT HadBC, and the T123A and A151V mutants to be less stable.

Besides initial velocities, enzyme stability was also tested using biophysical approaches. We analyzed the thermal stability of WT HadBC and the four resistant mutants using nano differential scanning fluorimetry (nanoDSF). HadC^V85I and HadC^K157R were the most stable constructs with inflection temperatures of 61.6 and 62°C, respectively, while HadC^T123A and HadC^A151V were the least stable at 60.7 and 60.8°C, respectively [Fig. 6B and Fig. S4]. These subtle changes in thermal stability indicate that the corresponding mutations alter the intramolecular interactions within the protein which could contribute to the observed differences in enzymatic activity in vitro [Fig. 6A] and in whole cells (see next section).

HadBC^V85I and HadBC^K157R can compensate for a deficiency in HadAB in Mtb –

The hypothesis that the HadBC mutant enzymes may compensate for a deficiency in HadAB was next tested genetically. To this end, attempts were made to disrupt by allelic replacement the entire hadABC operon of Mtb in the background of merodiploid strains expressing either WT versions of hadABC, hadAB and hadBC, or each of the four mutated hadBC variants from integrative plasmids. Double crossover (DCO) mutants were generated by recombineering. While the deletion of the hadABC locus was easily achieved in the backgrounds of merodiploid strains expressing extra-copies of hadAB (3 out of the 3 candidate mutants tested were DCOs) or hadABC (4 out of the 4 candidate mutants tested were DCOs) [Fig. 7A], DCO couldn’t be achieved, as expected, in the background of the merodiploid strain expressing an extra-copy of WT hadBC. No DCO mutants were obtained either in the merodiploid strains expressing the hadBC^A151V and hadBC^T123A variants. Deletion of the hadABC locus, however, was successful in the merodiploid strains expressing the hadBC^V85I and hadBC^K157R variants, with the greatest number of candidates consistently being obtained in the hadBC^K157R-expressing strain (5 out of the 11 candidate mutants analyzed were DCOs in the case of HadBC^K157R in one experiment compared to 2 out of 43 in the case of HadBC^V85I) [Fig. 7A]. The growth curves of the different rescued mutants in 7H9-OADC-Tyloxapol broth at 37°C are presented in Fig. 7B. Attempts to achieve DCO in the hadBC^A151V and hadBC^T123A merodiploid backgrounds by lowering the temperature at which the mutants were isolated from to 37°C to 32°C were unsuccessful.

Figure 7: Allelic replacement at the hadABC locus of Mtb rescued with WT hadAB or the hadBC variants hadBC^V85I and hadBC^K157R.

(A) Allelic replacement was confirmed by PCR using sets of primers located outside the linear hadABC allelic exchange substrate in one to two independent clones from each of the merodiploid strains. Different sets of primers were used for the various merodiploids. The WT 3.2 Kb amplification signal is replaced by a 4.1 Kb fragment in the mutants rescued with the two hadBC variants due to the replacement of 1.3 Kb of the hadABC locus by a 2.2 Kb-streptomycin resistance cassette. The WT 2.3 Kb amplification signal is replaced by a 3.2 Kb fragment in the mutants rescued with hadAB or hadABC due to the replacement of 1.3 Kb of the hadABC locus by a 2.2 Kb-streptomycin resistance cassette.

(B) Growth curve of the rescued mutants in 7H9-OADC-tyloxapol broth at 37°C.

(C) Effect of ISO treatment on the mycolic acid composition of Mtb hadABC knock-out mutants rescued with WT hadAB or the hadBC variants, hadBC^V85I and hadBC^K157R. Cells were either untreated or treated with the indicated concentrations of ISO, metabolically-labeled and processed as described in Figure 2A. The same volume of [¹⁴C]-acetate-labeled fatty acid and mycolic acid methyl esters (FAMEs and MAMEs) from treated and untreated cells were analyzed by TLC.

The Mtb hadABC knock-out mutant rescued with hadBC^V85I and hadBC^K157R proved to be fully protected from the effect of ISO and TAC on mycolic acid biosynthesis [Fig. 7C] consistent with their high-level resistance to both drugs [Table 2]. In sharp contrast, the Mtb hadABC mutant rescued with hadAB was hypersusceptible to the effects of the drugs.

Table 2: MIC of ISO and TAC against Mtb hadABC knock-out mutants rescued with WT hadAB or the hadBC V85I and K157R mutated variants.

MIC values are in μg/mL. MIC determinations were performed two to three times on two independent clones for each knock-out mutant. RIF: rifampicin; EMB: ethambutol. nd, not determined.

Strains	ISO	TAC	INH	RIF	EMB
Mtb H37Rv mc26206 (WT)	1.25	0.62	0.03	0.03	5
MtbΔhadABC/pNIP40b-hadAB	0.16	0.08	0.03	nd	2.5
MtbΔhadABC/pNIP40b-hadBCV85I	> 20	> 20	0.03	0.03	10
MtbΔhadABC/pNIP40b-hadBCK157R	> 20	> 20	0.03	0.03	5

ISO and TAC do not inhibit HadBC –

The predicted resemblance of the active sites of HadAB and HadBC finally led us to determine whether WT HadBC may serve as a target for ISO and TAC. The first indication that HadBC was not in fact a target was the lack of build-up of any form of 3-hydroxy meromycolate precursors in the drug-treated Mtb isogenic mutant expressing the HadA^C61S variant [Fig. 2B], which is the opposite of what would be expected if HadBC were inhibited. The lack of inhibition of the WT and mutant forms of HadBC by either the activated or non-activated forms of ISO and TAC in vitro and in whole cells was further reflected in enzyme inhibition assays [Fig. S5A], and the high-level resistance of the Mtb hadABC knock-out mutant rescued with hadBC^V85I and hadBC^K157R to both drugs [Table 2].

Further, direct binding studies of ISO and TAC with WT and mutant forms of HadBC via nanoDSF failed to reveal any evidence of binding for the two pro-drugs despite all forms of the protein demonstrating substantial thermal stabilization (+ 3.5 to 5.1°C) upon binding to trans-2-C_16:1-CoA [Fig. S5B]. Collectively, these results exclude the binding of HadBC mutants to ISO and TAC as a mechanism through which the HadC resistance mutations may titrate out the drugs and protect HadAB.

DISCUSSION

The dispensability of HadBC for Mtb growth and apparent lack of inhibition of this enzyme by ISO and TAC raised important questions as to the involvement of this enzyme in the resistance of Mtb to both drugs. The finding that HadBC enzymes harboring K157R and V85I mutations in the HadC subunit can rescue the viability of an Mtb hadABC knock-out mutant and render the bacilli immune to the effect of the drugs provides a rational explanation to the resistance observed in spontaneous HadC^K157R and HadC^V85I mutants. The ability of these HadBC mutants to compensate for the activity of HadAB further highlights the importance of developing inhibitors capable of inhibiting both dehydratases simultaneously, a goal that may be achievable given the structural resemblance of the two enzymes and their reliance on the same catalytic subunit, HadB. Structural modeling and molecular dynamic simulations indicated the K157R and V85I mutations are likely to alter substrate recognition, an assumption supported by our genetic rescue experiments and the results of our enzymatic assays comparing the activity of the HadC^K157R and HadC^V85I mutants to that of wild-type HadBC on an acyl-CoA substrate. A somewhat surprising observation from our studies, however, is the ability of the HadBC^K157R and HadBC^V85I enzymes to compensate for a deficiency in HadAB in Mtb despite their relatively low activity in comparison to HadAB under the conditions of our in vitro assay. If the activities of the HadBC^K157R and HadBC^V85I enzymes measured in vitro on a trans-2-C_16:1-CoA substrate reflect the actual activities of these enzymes in intact bacilli, this entails that Mtb can grow and sustain normal mycolic acid biosynthesis with only 1.5 to 2.5% of HadAB’s activity which argues against the vulnerability of HadAB and contrasts with the demonstrated potency of ISO and TAC in vitro and in vivo. A likely explanation to this apparent discrepancy is that the HadBC^K157R and HadBC^V85I variants display, as a result of their enhanced affinity for shorter substrates and increased stability [Fig. 6A and 6B], an activity in fact qualitatively and quantitatively much closer to that of HadAB in intact bacilli. Several reasons could account for the inability of our enzymatic assays to accurately capture this activity including their inability to measure the functional impact of the mutant enzymes’ increased stability, the relatively short length (C₁₆) of the substrate used in our assay compared to what HadAB is thought to process in the context of FAS-II (C₁₆-C₅₆)^19–20, the fact that we used acyl-CoA rather than the natural acyl-ACP substrates of these enzymes to measure enzyme activities, the fact that the activities measured are for the hydration rather than the dehydration reaction normally performed by the dehydratases in intact cells, and the possibility that HadBC may overall display a greater activity in the context of the FAS-II interactome than when tested as an isolated enzyme in vitro.

Despite HadBC^T123A exhibiting a comparable activity to that of HadBC^V85I and HadBC^K157R on trans-2-C_16:1-CoA in vitro, the HadBC^T123A variant failed to rescue the growth of MtbΔhadABC and so did the HadBC^A151V mutant, indicating that the HadAB-like dehydratase activity of these mutated HadBC enzymes is below the level required to sustain growth in the hadABC knock-out background. Incidentally, these two mutants are the ones whose stability was found to be decreased compared to the WT form of the enzyme in our thermal stability assay. Although this observation does not exclude a contribution of the increased HadAB-like activity of HadBC^T123A to the resistance of Mtb to ISO and TAC, it strongly suggests that other resistance mechanisms related to the T123A and A151V mutations are at play. We propose that perturbations caused by missense mutations in HadC on the FAS-II interactome contribute - to an extent dependent on the nature of each mutation - to the resistance of Mtb to ISO and TAC by limiting the drugs’ access to HadAB. This assumption is based on two main observations. First is the fact that the isogenic HadC^T123A resistant mutant and, to a lesser extent, the three other mutants display mycolic acid profiles consistent with alterations in the activity of enzymes functionalizing the meromycolate chain [Fig. 3A–B]. Since both HadAB and HadBC have been reported to physically interact with some of the same mycolic acid S-adenosyl-methionine-dependent methyltransferases¹⁷, it is conceivable that the activity of these enzymes in particular, and the accessibility of HadAB to ligands may be impacted as a result of conformational changes in HadBC. Second is the dramatically increased (8- to more than 40-fold) susceptibility of Mtb hadC deletion mutants to ISO and TAC¹¹, whereas a ~10 to 14-fold increase in the expression of hadBC results, on the contrary, in high-level resistance to both drugs^10–13. These observations suggest that the physical presence of the HadC subunit makes HadAB less vulnerable to the action of the drugs. Clearly, more work is required to elucidate the impact of HadC on the FAS-II interactome and accessibility of HadAB to ISO and TAC. Given the complexity of this interactome and likely involvement of multiple proteins in the protection of HadAB from the drugs, such studies will require a detailed analysis of the composition and structure of the entire multiprotein complex centered on FAS-II in WT and mutant hadC-expressing strains.

METHODS

Bacterial strains and culture conditions –

The avirulent auxotrophic Mtb H37Rv strain mc²6206 (ΔpanCDΔleuCD) was grown at 37°C in Middlebrook 7H9-OADC-0.05% tyloxapol supplemented with 0.2% casaminoacids, 48 μg/mL pantothenate and 50 μg/mL L-leucine or on similarly supplemented Middlebrook 7H11-OADC agar medium. E. coli BL21 (DE3) was grown in Luria Bertani (LB) medium (10 g/L Bacto-tryptone, 5 g/L yeast extract and 5 g/L NaCl) (Difco) and on LB agar at 37°C. Kanamycin (Kan; 25 μg/mL), streptomycin (Str, 20 μg/mL), hygromycin (Hyg; 50 μg/mL) and ampicillin (Amp; 50 μg/mL) were added as needed.

MIC determinations –

The susceptibility of the various Mtb recombinant strains to antibiotics was determined in 96-well microtiter plates at 37°C in 7H9-ADC-0.05% Tween 80 medium using the resazurin blue test²⁵.

Construction of isogenic Mtb strains expressing mutated forms of HadC and disruption of hadABC –

Recombineering was used to introduce the V85I, T123A, A151V and K157R missense mutations in HadC of Mtb H37Rv mc²6206 by allelic replacement. To this end, the Gp60 and Gp61 recombineering proteins from mycobacteriophage Che9c were expressed in Mtb H37Rv mc²6206 from the replicative plasmid pJV53-XylE under control of an acetamide-inducible promoter^26–27. Electrocompetent cells of Mtb H37Rv mc²6206/pJV53-XylE were prepared upon induction of the recombinase with acetamide and electrotransformed using 1 μg of hadC DNA oligomers to introduce the SNPs [Table S1]. Following transformations, cell suspensions were plated on 7H11-OADC plates containing 5 to 10 μg/mL ISO or 1 to 2 μg/mL TAC and incubated at 37°C for 3–4 weeks. Resistant colonies were cultured in liquid medium and the presence of SNP was verified by PCR and sequencing of the entire hadC gene.

pNIP40b-hadABC, pNIP40b-hadAB, pNIP40b-hadBC^WT, pNIP40b-hadBC^V85I, pNIP40b-hadBC^T123A, pNIP40b-hadBC^A151V and pNIP40b-hadBC^K157R, the integrative plasmids used to rescue the hadABC knock-out mutant, were constructed by cloning the hadABC, hadAB and five different hadBC variants expressed from the hsp60 promoter in the pNIP40b plasmid²⁸. The linear allelic substrate used to delete the hadABC locus was generated by bracketing the streptomycin-resistance cassette from pHP45Ω with 963 bp (including the first 160-bp of hadA) and 960 bp (including the last 150-bp of hadC) of upstream and downstream DNA sequence flanking the hadABC operon. Acetamide-induced Mtb H37Rv mc²6206 cells harboring pJV53-XylE and each of the seven pNIP40b integrative plasmids described above were electro-transformed with this linear substrate and double-crossover mutants were selected on Str- and Hyg-containing medium. Acetamide-induced Mtb H37Rv mc²6206 cells harboring pJV53-XylE and an empty pNIP40b plasmid were similarly transformed and double crossover mutants selected as above.

Whole cell radiolabeling experiments –

Radiolabeling of Mtb WT and resistant mutants (0.5 μCi/mL; specific activity, 52 Ci/mol, Perkin Elmer) was performed for 13 hr at 37°C with shaking. [1,2-¹⁴C]acetic acid was added to the cultures at the same time as ISO and TAC. Preparation of fatty acid and mycolic acid methyl esters from whole cells followed earlier procedures¹¹. [1,2-¹⁴C]acetic acid-derived fatty acid and mycolic acid methyl esters were separated by TLC on aluminum-backed silica gel 60-precoated plates F₂₅₄ (E. Merck) and revealed by PhosphorImaging.

LC/MS and GC/MS analyses –

Analyses of mycolic acids by LC/MS and meromycolate precursors by LC/MS and GC/MS were conducted as described by Grzegorzewicz et al.¹¹.

RNA preparation, reverse transcription and qRT-PCR –

Mtb RNA was extracted from 5-mL cultures grown to an OD₆₀₀ of 0.2 using the Direct-zol^™ RNA Miniprep kit (Zymo Research) per the manufacturer’s instructions. Reverse transcription reactions were carried out using the Superscript IV kit (Invitrogen), and qRT-PCRs were run using the SsoAdvanced^™ Universal SYBR® Green Supermix kit (Bio-Rad) as per the manufacturers’ protocols and analyzed on a CFX96 real-time PCR machine (BioRad). PCR conditions: 98°C (2 min; enzyme activation), followed by 39 cycles of 98°C (5 sec; denaturation) and 55°C (5 sec; annealing/extension). Mock reactions (no reverse transcription) were done on each RNA sample to rule out DNA contamination. The target cDNA was normalized internally to the sigA cDNA levels in the same sample. The primer sequences are provided in Table S1.

Dehydratase assays –

HadAB and HadBC (including WT and mutated variants V85I, T123A, A151V and K157R) production and purification, and dehydratase assays followed the procedures described by Sacco and collaborators¹⁹.

Synthesis of trans-2-hexadecenoyl-CoA –

trans-2-hexadecenoyl-CoA (trans-2-C_16:1-CoA) was synthesized by following the procedures described by Al-Arif and Blecher²⁹. Purity and identity were confirmed by ULPC-ELSD-MS analysis to be >95%.

Structural modeling and molecular dynamics simulation –

A homology model of the HadBC complex was built using the crystal structure of HadA²⁴ (PDB code 4RLT) as a template for HadC. The sequence identity was up to 45% and similarity was up to 64%. Molecular dynamics calculations were carried out with the AMBER14 program using the FF14SB force field. Parameters for monovalent ions were obtained from the ionsjc_tip3p set of parameters. The whole complex was solvated in an octahedron box of TIP3P water molecules and neutralized by adding sodium ions. Periodic boundary conditions, particle-mesh, Ewald treatment of the long-term electrostatics and SHAKE-enabled 2-fs time steps were employed. A two-stage energy minimization was performed, followed by a gradual heating of the complex system from 0 K to 300 K over 20 ps and a 50-ps equilibration. An additional 0.5-ns simulation at 300 K was performed to further optimize the system. All production runs were performed with the NPT ensemble for 1 μs.

The channel volumes were calculated using the online server Voss Volume Voxelator³⁰. The inner probe radius was set at 1.4Å and the outer probe radius was set at 6Å. In the case of the WT HadBC enzyme, the program found two adjoining channels instead of one. The volume of the two channels were added to arrive at the total volume.

Thermal stability and binding assays –

The thermal stability of the HadBC WT and mutant proteins (including variants V85I, T123A, A151V and K157R) was measured using a Nanotemper Tycho NT.6, which measures intrinsic fluorescence of a protein at 330 nm and 350 nm from 35°C to 95°C, taking a reading approximately every 0.1 degrees. For the stability testing, purified protein was loaded into the Tycho capillaries and analyzed in triplicate. For the binding studies, samples (in a total volume of 40 μL) were prepared in 5% DMSO, with a total protein concentration of 40 μM and a final ligand concentration of 500 μM. These samples were also analyzed in triplicate. The data was analyzed using the internal software from the Tycho instrument to obtain the first derivative of the fluorescence intensity at 330 nm. This first derivative data was then imported into GraphPad Prism for error calculations, normalization for comparison, and plotting of the graphs.

Abbreviations

DCO

double cross-over

FAMEs

fatty acyl methyl ester

ISO

Isoxyl

MAMEs

mycolic acid methyl ester

Mtb

Mycobacterium tuberculosis

TAC

Thiacetazone